Multilayer transient testing commonly relies on a series of step rate changes in surface flow rate with acquisition of stabilized rate profiles before each rate change and stationary downhole transient rate and pressure measurements after each rate change. The procedure requires one rate change for each layer to be characterized, and overall test duration can be quite long depending on the number of layers to be characterized. This study introduces an entirely new testing approach that uses transient temperature data at multiple locations together with a single-point transient pressure measurement. Because the temperature sensors acquire multipoint temperature simultaneously, this technique requires only one surface flow rate changes, thereby reducing the test duration significantly.

A coupled wellbore/reservoir thermal model developed in a previous study showed that the combination of transient temperature and pressure is sufficiently sensitive to individual layer properties to determine layer permeability and skin values in multilayered systems. In this work, the inverse problem is solved using the Levenberg-Marquardt regression algorithm.

In this new testing method, temperature data from only n+1 locations are required to determine layer permeability and skin values in an n-layer reservoir, and strategic sampling over time accelerates the regression convergence. The inversion has been tested on many synthetic cases with wide variations in layer permeability and skin values and has proven to be both accurate and robust. In addition, temperature resolution and data signal to noise impacts have been studied along with a data filtering approach that enable selection of suitable pressure and temperature sensor technologies for applying the new testing method. With the interpreted permeability and skin values, well productivity can then be enhanced by hydraulically fracturing in low permeability layers and/or by applying matrix stimulation in layers with high skin.

You can access this article if you purchase or spend a download.